← Back to Mycal Terms

Low Entropy Attractor

First used: March 2026

Definition

A region of embedding space that an AI inference system repeatedly collapses into under query pressure — structured, retrievable, and statistically efficient to reuse. A phrase or framework becomes a low entropy attractor when it simultaneously compresses more meaning than its length suggests, is reinforced across multiple document types and structural contexts, and sits in a topologically dense neighborhood of cross-linked, canonically anchored definitions. Unlike thermodynamic attractors, which form through natural dissipation, low entropy attractors in language systems can be engineered through structured repetition, canonical identifiers, and deliberate semantic reinforcement across a document network.

Sources